Team-based collaboration is integral to education, work, and daily life, fostering ability-driven peer effects through discussions, social comparisons, and knowledge sharing. Despite extensive evidence of peer effects in specific contexts, their broader impacts on comparable but different activities remain underexplored. Our study addresses this gap using a novel dataset from Mongolia that combines a natural field experiment in classrooms, university entrance examination scores, and grade point averages in the university. First-year undergraduate students were randomly paired to collaboratively complete weekly assignments throughout a course. Low-ability students (based on their entrance exam scores) paired with high-ability peers significantly improved their academic performance not only in the specific course but also in other concurrent courses, showing strong spillover effects. The magnitude of the spill-over relative to the direct effect was 0.723. These pairings had no adverse effects on high-ability students. The findings highlight the Pareto efficiency of peer interactions in groups with large ability differences and offer insights into improving productivity and learning through ability-based spillovers.